Inference and checking of object ownership

Abstract. Ownership type systems describe a heap topology and enforce an encapsulation discipline; they aid in various program correctness and understanding tasks. However, the annotation overhead of ownership type systems has hindered their widespread use. We present a unified framework for specification, type inference and type checking of ownership type systems, and instantiate the framework for two such systems: Universe Types and Ownership Types. We present an objective metric defining a “best typing ” for these type systems, and develop an inference approach that maximizes the metric. The programmer can influence the inference by adding partial annotations to the program. We implemented the approach on top of the Checker Framework and present the results of an experimental evaluation.